Researching the web
Agent Skills for using Scout: web research, sourced answers, company enrichment, structured extraction, monitoring. Works in Claude Code, Claude.ai, the Agent SDK, and Vercel eve.
npx -y skills add Scout-AI-Labs/scout-skills --skill researching-the-webAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Research a question using the live web with Scout. Use when the user asks you to research a topic, gather sources, fact-check a claim, write a briefing or report from current information, or answer something your training data can't cover on its own. Plans sub-questions, searches and reads real pages, and writes a cited answer.
SKILL.md
1.9 KB, 405 tokens by cl100k_base, as published. Nobody here has run it
Researching the web with Scout
Use this when a question needs current, sourced information rather than recall. Scout handles live search and reliable scraping; you handle the reasoning.
How to reach Scout
- If the Scout MCP server is connected, use its tools:
scout_search,scout_scrape,scout_extract. - Otherwise call the API directly with
curl(the key is inSCOUT_API_KEY):
curl -s https://core.usescout.sh/v1/search \
-H "Authorization: Bearer $SCOUT_API_KEY" -H "Content-Type: application/json" \
-d '{"queries":["<query>"],"limit":6}'
curl -s https://core.usescout.sh/v1/page/markdown \
-H "Authorization: Bearer $SCOUT_API_KEY" -H "Content-Type: application/json" \
-d '{"url":"<url>","max_chars":6000}'
Workflow
- Plan. Break the question into 3-5 focused sub-questions. Cover the distinct angles, don't overlap.
- Search. Run each sub-question through search. For a hard or broad question, set
"depth":"deep"so Scout runs a multi-step search itself. - Read. Scrape the 2-3 most relevant results per sub-question to Markdown. Prefer primary sources over aggregators.
- Synthesize. Write the answer grounded only in what you read. Cite each claim with the source URL. Where sources disagree or you came up short, say so instead of guessing.
Notes
- Keep a running list of the URLs you actually used; put them at the end as a numbered Sources list.
- If a page fails to scrape, skip it and note the gap rather than inventing the content.
- Don't pad. A tight, sourced answer beats a long one.